Temperature problems, Xeon E5-2690 v4 no thermal throttling

[blkopler]

 

Hello!

 

Everytime I integrate several scenes inside stash (https://github.com/stashapp/stash) my CPU temperature rises over 100 °C and slowly climbs higher until stash is done. The docker container uses CPU core 0-7 (and HT 14-21) up to 100% utilization.

Shouldn't the CPU throttle when it gets to hot? Is it somehow possible to activate thermal throttling?

The CPU fan seems to get louder, the case fans however don't seem to change, hard to judge.

Is there a way to create a load, to stress test the CPU in unraid, so I can reproduce the temperature increase?

 

Any good CPU cooler upgrade that is tolerable quiet, the NH-U12DX i4 doesn't seem to fit (3mm argh) and 2011-3 narrow ILM makes it hard to find something.

 

Mainboard: Asrock Rack EPC612D8A  (https://www.asrockrack.com/general/productdetail.asp?Model=EPC612D8A#Specifications)

CPU: Intel® Xeon® CPU E5-2690 v4 @ 2.60GHz (https://www.intel.com/content/www/us/en/products/sku/91770/intel-xeon-processor-e52690-v4-35m-cache-2-60-ghz/specifications.html)

CPU Cooler: Noctua NH-D9DX i4 3U (https://noctua.at/en/nh-d9dx-i4-3u)

Chassis: Inter-Tech 4416 (https://www.inter-tech.de/en/products/ipc/storage-cases/4u-4416)

Unraid: 6.10.0-rc4

Stash: 0.15.0

Fans: 

1. Noctua NF-R8 redux-1800 PWM (https://noctua.at/en/nf-r8-redux-1800-pwm)
2. Noctua NF-R8 redux-1800 PWM (1 & 2 together via 4-Pin Y-Cable [https://noctua.at/en/na-syc1] on the same fan header)
3. Noctua NF-S12B redux-1200 PWM (https://noctua.at/en/nf-s12b-redux-1200-pwm)
4. Noctua NF-S12B redux-1200 PWM
5. Noctua NF-S12B redux-1200 PWM

 

(In use the chassis is closed.)

waterline-sv-diagnostics-20220602-0342.zip

[MAM59]

this looks pretty much cool enough for this old and slow cpu.

Are you sure that the cpu cooler has good contact (thermal paste?) to the cpu?

Or are you running the box in the sahara?

Or have you turned down all fans too much in the Bios? (try to give them 10% more minimum rotation or a steeper rising curve)

 

Does it even overheat with case open (aka: is the airflow from the front somehow blocked?)

 

100° should never be reachable with this equipment. I would go for the thermal paste first. too few is bad and too much too. The Xeon is a very simple cpu compared to modern types, you just need one drop in the very middle of the chip.

 

 

[JonathanM]

When the CPU is reporting high temps, is the heatsink too hot to touch? If yes, airflow is the issue. if not, heatsink compound / mating surface flatness / mounting tension is the issue.

[blkopler]

2 hours ago, JonathanM said:

When the CPU is reporting high temps, is the heatsink too hot to touch? If yes, airflow is the issue. if not, heatsink compound / mating surface flatness / mounting tension is the issue.

 

At 57°C the heat pipes and the fins feel hot to the touch.

 

12 hours ago, MAM59 said:

this looks pretty much cool enough for this old and slow cpu.

Are you sure that the cpu cooler has good contact (thermal paste?) to the cpu?

Or are you running the box in the sahara?

Or have you turned down all fans too much in the Bios? (try to give them 10% more minimum rotation or a steeper rising curve)

 

Does it even overheat with case open (aka: is the airflow from the front somehow blocked?)

 

100° should never be reachable with this equipment. I would go for the thermal paste first. too few is bad and too much too. The Xeon is a very simple cpu compared to modern types, you just need one drop in the very middle of the chip.

 

 

 

The room temperature is at 20-30°C, and there are is enough space in front and back of the case.

I checked the bios and all fans were on [Smart Fan] (100% at 78°C), as the CPU fan seems to increase at load, I put all the case fans on [Full On]. It doesn't feel like they are on 100%.

If I remember correctly than opening the case didn't effect the temperature much (+- 1-2°C).

 

Using an online stress test (https://cpux.net/cpu-stress-test-online) in a VM:

Closing the case increased the temperature from 69°C to 77°C. The case fans didn't seem to increase in speed, holding a small paper at the fan, the paper moves 10° maybe 15° angle. Hm, the VM uses 12 HTs (see picture) and does not increase the temperature in such a rapid way. Stash on the other hand uses 16 HTs (see the grey ones in the picture) and reaches critical temperature fast.

It seems the case fans don't work as they should.

 

Tried using pwmconfig: 

Found the following devices:
   hwmon0 is coretemp
   hwmon1 is nct6776

Found the following PWM controls:
   hwmon1/pwm1           current value: 127
   hwmon1/pwm2           current value: 127

Giving the fans some time to reach full speed...
Found the following fan sensors:
   hwmon1/fan1_input     current speed: 0 ... skipping!
   hwmon1/fan2_input     current speed: 0 ... skipping!

There are no working fan sensors, all readings are 0.
Make sure you have a 3-wire fan connected.
You may also need to increase the fan divisors.

 

So what can I do? Should I buy one of those fancy old school fan controller?

[blkopler]

What is strange is, that unraid will notify me if disks are to hot, but my CPU can melt for all it cares. I only stumbled over it by chance. And shouldn't there be a blue screen if the temperarture reaches a critical level?

[blkopler]

So, the problem seems to be, that the case fans stayed idle, like said before the fans were on [Smart Fan] and should have increased with the temperature. The CPU fan did, the rear and front ones didn't. Maybe they took the motherboard temperature as measure? Anyway, bought myself a fan hub (https://www.arctic.de/en/Case-Fan-Hub/ACFAN00175A) and used the CPU fan as indicator and now the case fans seem to rise in speed accordingly.